The Agent Skills Directory

[COMMAND_EXECUTION]: The terminal_run and terminal_send_keys tools provide the ability to execute arbitrary shell commands on the host system within persistent tmux sessions. This core functionality allows the agent to perform any action the user has permissions for.
[EXTERNAL_DOWNLOADS]: The terminal_prepare_skill_prerequisites tool is designed to automate the installation of third-party command-line interfaces (CLIs) and manage authentication flows. This involves downloading and executing external code at runtime to set up tools like gemini-cli or railway.
[DATA_EXFILTRATION]: The terminal_read tool allows the agent to capture the current visible output of any terminal window. This presents a risk of sensitive data exposure if credentials, environment variables, or private file contents are displayed in the terminal buffer during a session.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from command outputs.
Ingestion points: Untrusted data from the host system or network services enters the agent's context through terminal_read and terminal_run tools defined in SKILL.md.
Boundary markers: The skill documentation does not specify the use of delimiters or instructions for the agent to ignore potentially malicious commands embedded in terminal output.
Capability inventory: The skill possesses high-impact capabilities including arbitrary command execution via terminal_run and terminal_send_keys, and session manipulation via terminal_kill_session in SKILL.md.
Sanitization: No sanitization or validation logic is mentioned to prevent the agent from being influenced by instructions contained within processed terminal data.

terminal